RE: JSON-LD requirements

>> I think since we are working towards Linked Data in JSON we
>> should leverage JSONs structural elements and as thus I really
>> think we should stop concentrating on triples.
> If you stop thinking about triples (which isn't an RDF invention)
> you are basically throwing out the baby with the bath water re.
> the fundamental goal.
>
> The goal is to use triples a mechanism for "data representation" via
> EAV/SPO graphs. 

What is the fundamental goal? If we aren't going to use the fundamental features of JSON I'm not sure why we are basing our approach on JSON at all. What's the difference to, e.g., Turtle then?

I agree that objects (subjects, nodes, however you would call them), attributes and, if possible, attribute values should be identified by IRIs that resolve to a description about what they are about. So in my opinion the challenge is how to map the various constructs in a JSON representation to IRIs in the simplest possible way. I would like to be able to describe arbitrary (not sure if we can achieve that) JSON so that a machine client can "understand" it.

This would us then allow to create generic clients for processing, querying and creating data. Since JSON is (mostly) used in Web APIs I would like to see that happen with Web APIs in mind.

Received on Monday, 4 July 2011 13:06:08 UTC